International Society for Soil Mechanics and Geotechnical Engineering

INTERNATIONAL SOCIETY FOR SOIL MECHANICS AND GEOTECHNICAL ENGINEERING

This paper was downloaded from the Online Library of the International Society for Soil Mechanics and Geotechnical Engineering (ISSMGE). The library is available here: https://www.issmge.org/publications/online-library

This is an open-access database that archives thousands of papers published under the Auspices of the ISSMGE and maintained by the Innovation and Development Committee of ISSMGE.

The paper was published in the proceedings of the 17th African Regional Conference on Soil Mechanics and Geotechnical Engineering and was edited by Prof. Denis Kalumba. The conference was held in Cape Town, South Africa, on October 07-09 2019.

Proceedings of the 17th African Regional Conference on Soil Mechanics and Geotechnical Engineering. 7, 8 & 9 October 2019 – Cape Town The limitations of standpipe piezometers in stability analysis

L. Geldenhuys, Y. Narainsamy & F. Hörtkorn Jones & Wagener, Johannesburg, South Africa

ABSTRACT: The pore pressure regime, material strength parameters and geometry of a Tailings Storage Fa- cility (TSF) influence the stability, often expressed as a Factor of Safety (FoS). Of these variables it is the pore pressure that is the likely to change in temporal and spatial variance within a relatively short period of time and which will have a significant impact on the FoS of the facility. Standpipe piezometers are often used in South Africa to monitor pore pressures within a TSF. Analyses were conducted to compare the FoS calculated for non-hydrostatic spatial distributions of pore pressures and the FoS calculated for sketched phreatic levels and hydrostatic conditions. Phreatic levels were from water levels in hypothetic standpipe piezometers that would measure pore pressures in the reference model. The FoS values calculated showed that the FoS for the models with pore pressures from the sketched phreatic levels under hydrostatic conditions were non-conservative. This highlights the need for alternative methods, such as regular piezocone probing, to determine the pore pressures in a TSF for stability monitoring.

1 INTRODUCTION Of these, it is the pore pressure regime that is most likely to vary rapidly throughout the lifetime of the Tailings Storage Facilities (TSFs) are used to store TSF and will affect a change in the Factor of Safety mine waste residue. The properties of this waste ma- (FoS) against failure. This paper aims to highlight the terial are generally related to the parent rock from influence that estimating the phreatic level from which it is being mined as well as the process method standpipe piezometers and assuming hydrostatic con- used to extract minerals. If the production process ditions can have on the calculated FoS compared to does not change, and the material is from a single when the FoS is calculated with the actual pore pres- plant, it is unlikely that the tailings material properties sures that are non-hydrostatic. The scope is limited to will change in a short period of time. an idealized gold TSF in which the material proper- Irrespective, the material properties will vary spa- ties are homogeneous. tially within the TSF due to segregation during deposition. Piezocone testing has been proven to be an effective method to determine the material properties in 2 PORE PRESSURES WITHIN A TSF the TSF. Hydraulic deposition is the primary deposition The seepage and pore pressure regime within a TSF method used in South Africa. The residue is deposited can have spatial and temporal variation and is in slurry form and, as it settles and consolidates, ex- influenced by deposition cycles and deposition cess water is returned to the plant to be re-used. This control, rainfall, decanting procedures, drainage is predominantly through decanting structures. A net- conditions, facility height, base geometry, work of drains can be implemented to collect some of consolidation, etc. Recently, the increase in the the interstitial water. The interstitial water, the expul- number of lined facilities, due to the promulgation of sion of excess water due to consolidation and the as- more stringent environmental legislation in South sociated flow to the drains or decant structure result Africa, has resulted in the assumptions regarding in a pore pressure distribution within the TSF that drainage conditions having to be reconsidered. The may vary spatially. flow conditions are three-dimensional, temporally The material properties and spatial distribution, variable and are therefore complicated to predict. TSF geometry as well as the pore pressure distribu- Pore pressures are often measured in a TSF as an tion within a TSF control the stability of the TSF input parameter to conduct regular stability analyses. (Wagener et al. 1998). The pore pressures can be determined by conducting

435 17th ARC Conference 2019 Cone Penetration Testing with pore pressure (considering the flow regime) while the dashed line measurements (CPTu) probing, also referred to as represents the phreatic surface inferred from the piezocone probing. piezometer measurements without considering the Although this provides an accurate representation flow regime. Note that, although the pore pressures of the pore pressure in the TSF (Wagener et al., 1998), indicated by the piezometers are correct, the inferred the costs associated with conducting these tests and phreatic surface is below the true phreatic surface. the fact that the results are only representative for the time (therefore also the dam geometry) at which probing was done warrants that alternative solutions are required to measure pore pressures at shorter intervals. In order to obtain continuous data for monitoring, permanent fixture devices are installed to indicate pore pressures in a TSF. In South Africa, the pore pressures are traditionally measured using open-end standpipe piezometers and vibrating wire piezometers. Both measure pore pressure at a specific depth below the surface and, to effectively use the measured data, the reference elevation of these depths needs to be known. Other devices, such as twin-tube hydraulic piezometers, porous piezometers, Figure 1. Difference in true phreatic level and piezometer level pneumatic piezometers and electrical resistance (Blight 2010) piezometers (among others) are available to the industry (Ridley et al. 2003). Although available, not all these methods are implemented. 3 METHOD AND ANALYSIS The open-end standpipe piezometer is the most commonly used in the experience of the authors and 3.1 Scope consists of a filter at the end of tube or pipe that is The scope of this paper is limited to the comparison extended to the surface. The piezometer will normally of the FoS calculated with phreatic levels inferred be installed in the centre of a drill hole, the end filled from standpipe piezometers and hydrostatic with wet sand (adjacent to the porous filter) and the conditions and the FoS calculated with pore pressures remainder of the gap around the pipe (above the from CPTu testing for stability analysis on a TSF. The porous filter) will be grouted closed. Because of this analysis was conducted by calculating the FoS using installation technique, the water level in the standpipe limit equilibrium slope stability analysis (method of only responds to the water pressure at the bottom of slices according to Morgenstern-Price, 1965) for the the standpipe and is isolated from other pore pressure outer wall of an idealised TSF with typical material regimes along the length of the standpipe. The water parameters for gold tailings. The material parameters level in the standpipe is therefore a measure of the used are summarised in Table 1. equipotential at the tip and, as such, the water level in The model geometry and phreatic lines considered a standpipe will rise to the potential at the bottom of are shown in Figure 2 and are based on typical TSFs the standpipe. assessed by the authors. The assignment of the mate- The seepage regime within a TSF seldomly results rial regions as well as the bench dimensions are in hydrostatic pore pressure conditions, particularly shown in Figure 3. The analysis was repeated for the towards the outer wall of the facility where flow lines case of a starter wall. This was done to force the slip are directed to underdrains. A flow gradient of surfaces to be above the bottom of the standpipes, 7 kPa/m has been considered to indicate drainage is based on the assumption that the starter wall is con- occurring (Vermeulen 2001) at that section and/or a structed with a material with a higher shear strength permeable foundation. Due to non-hydrostatic than the surrounding tailings. conditions there is often a difference between the water level within a standpipe piezometer and the true Table 1. Summary of material parameters phreatic level in the region where the standpipe is Material Unit weight Friction Cohesion installed. A piezometer will therefore give a false (kN/m3) angle (°) (kPa) phreatic level if the rate of pore pressure increase with Tailings 18 33 0 Base 22 40 0 depth is not hydrostatic. The magnitude of this Foundation 22 33 5 difference is associated to the depth to which a Starter wall 22 40 0 standpipe is installed and the rate of pore pressure increase with depth. This difference is depicted in Figure 1. The upper, solid line of the flownet represents the true phreatic surface within the TSF

436 L. Geldenhuys, Y. Narainsamy & F. Hörtkorn was done to have a representation of various failure mechanisms. These analyses formed a single analysis set. The analysis set (Case A, B and C for each of Case 1, 2 and 3) was then repeated for the model with a starter wall (Fig. 3). Once all the critical FoS values for the reference case were determined, the analysis was then repeated, only this time the pore water pressure in the model was derived by drawing in the phreatic surface and Figure 2. Model geometry (no starter wall region assigned) assigning hydrostatic conditions in the saturated zone below the phreatic surface. The phreatic level at each standpipe was determined by assuming that the pore pressure at the base of the standpipe would result in a hydrostatic water level in the standpipe. Standpipes were assumed to be installed to a depth of 1 m above the base layer. The method used to calculate the expected water height in the standpipe, and thus the phreatic level that was drawn in, is illustrated in Figure 4. The critical FoS was then calculated for this set as it was calculated for the set that was based on

the spatial pore pressure distribution. Figure 3. Material regions and bench dimensions (starter wall region assigned) For the case shown, the spatial distribution of pore pressures increases at a rate of 7 kPa/m with depth 3.2 Analysis matrix from the phreatic level. This would determine the The reference FoS was calculated from analysis in pore pressures around the piezometer for the which a spatial distribution of pore pressures (Table calculation of the reference FoS. For comparison, the 2) was modelled in the analysis. This was done by phreatic level would then be drawn in the analysis by defining pore pressures from Table 2 at the location calculating a hypothetic standpipe water level. The of each standpipe in the model. The pore pressures for standpipe water level would be calculated according the rest of the model were interpolated and to the pore pressure (from the reference case) at the extrapolated using the Kriging method. The Kriging tip of the standpipe and relating that to a water level method is a first order spline interpolation technique above that point for hydrostatic conditions. (Cressie 1990) and is used in the modelling software It should be noted that critical failure mechanisms to estimate the pore pressures between values entered were determined independently for each analysis set. at the piezometer locations. The comparison was therefore not for the same slip Three cases with different rates of pore pressure surface, but rather between critical slip surfaces. build-up at each piezometer were modelled for the These could be different in depth and geometry. This analysis (Case 1, 2 and 3), each with the phreatic level was done to represent what would typically occur if at a different elevation as shown in Figure 2. This analysis was conducted for piezometer-based phreatic determined the elevation of zero pore pressure and the levels without knowing the pore pressure distribution pore pressures below this elevation were determined within the TSF. and used in the model according to the rate of pore All other boundary conditions were kept the same pressure increase with depth at the location of the in the analysis. piezometer. For each case, three combinations of pore pressure distributions were used according to Table 2. For Case A, a higher rate of pore pressure build-up is assumed at the toe (simulating poorer drainage conditions towards the toe compared to the basin). For Case B, a higher rate of pore pressure build-up closer to the pool was assumed (simulating poorer drainage at the toe when compared to the basin) and for Case C a constant rate of pore pressure build-up from the toe to the pool was assumed (simulating functional underdrains). The critical FoS (i.e. the lowest FoS for the given possible slip mechanisms) was determined for a failure of the lowest bench, failure through the second bench and failure through the entire outer wall. This Figure 4. Method for calculating the water level in the standpipe 437 17th ARC Conference 2019 Table 2. Summary of variables Rate of pore pressure increase with Phreatic depth (kPa/m) line Piezo 1 Piezo 2 Piezo 3 Piezo 4 Case A 9 8 7 6 1, 2 & 3 Case B 6 7 8 9 1, 2 & 3 Case C 7.5 7.5 7.5 7.5 1, 2 & 3

4 RESULTS

The results are combined for each geometry type (starter wall and no starter wall) as well as for each failure location (first bench, second bench and entire outer wall). Figure 7. Failure of entire outer wall (no starter wall) Each column refers to a specific case. “Case 2B”, for example, refers to Phreatic Line 2 (Fig. 2) and pore pressure distribution Case B (Table 2). The “percentage difference” (see Eq. 1) refers to the percentage difference between the reference FoS (퐹표푆) (from analysis in which pore pressure distribution is assigned throughout the model) and the FoS (퐹표푆) from the analysis in which hydrostatic conditions are assumed below a phreatic level that is drawn in the model from the calculated standpipe water levels. 푃푒푟푐푒푛푡 푑푖푓푓 = ∗ 100 (1)

Figure 8. Failure of first bench only (with starter wall)

Figure 5. Failure of first bench only (no starter wall)

Figure 9. Failure up to second bench (with starter wall)

Figure 6. Failure up to second bench (no starter wall) Figure 10. Failure of entire outer wall (with starter wall)

438 L. Geldenhuys, Y. Narainsamy & F. Hörtkorn 5 DISCUSSION OF RESULTS Line 2). This could be due to the localised (in the region of the toe) difference in elevation between the When comparing the analysis in which the default true and the plotted phreatic level resulting in greater model geometry was used (Figs. 5, 6 & 7) and that in weight being applied to the slices on the driving side which the geometry included a starter wall (Figs. 8, 9 of the failure in the model with a spatial pore pressure & 10), it is evident that the difference in the FoS be- distribution than the model with the plotted phreatic tween the two methods of modelling pore pressures line. was more pronounced in the models with the starter wall. It was noticeable that the critical failure plane, regardless of the size, was higher above the base of the TSF for the models with the starter wall than for models without the starter wall. At the base of the facility, there would be very little to no difference in pore pressures for the two methods of defining the pore pressure regime. For a failure plane close to the tips of the piezometers (Fig. 11), the pore pressures at the base of the slices would be similar for both methods and therefore the effective stress would also be similar, hence the smaller difference in the FoS. Once Figure 11. Failure through base of TSF the failure plane is further from the base (Fig. 12), the pore pressure difference between the two methods would be greater resulting in a greater difference in effective stress at the base of the slices and a greater difference in FoS between the two methods. Overall, the results show a greater percentage difference for Phreatic Line 2 (Case 2) and the smallest percentage difference (sometimes no difference) for Phreatic Line 3 (Case 3). Since it was assumed that the water level in the piezometer would represent a pore pressure near the base of the tailings, a greater elevation difference between the reference phreatic Figure 12. Failure forced to be higher than the base of TSF level and the bottom of the standpipe would result in a greater difference between the reference phreatic level and the one plotted from the water levels in the 6 CONCLUSIONS standpipes. In other words, the underestimation of FoS by using piezometer water levels to model the The analysis conducted shows that the FoS calculated phreatic level becomes greater the higher the water when the true pore pressure distribution is used in a levels in the TSF rise, which is when the determina- slope stability model for a TSF outer wall is lower tion of the stability of the TSF becomes all the more than the FoS calculated when the water level in the important. standpipes is assumed to be the phreatic level and hy- In the case of a model with a starter wall (Figs. drostatic conditions are applied. This conclusion ap- 8, 9 & 10), the difference in the FoS was zero in some plies to the assumption made regarding the elevation cases. This is because, regardless of the size of the of the tip of the piezometers. This obviously does not failure, the critical failure mechanism did not inter- apply when the critical slip plane analysed does not cept the starter wall. This also resulted in the entire intercept the phreatic level at all. failure mechanism occurring above the phreatic sur- This is similar to the work done by Van der Berg face for both cases of modelling pore pressures. The (1995) in which it was found that stability analyses in differences in localised pore pressures therefore had which phreatic levels were plotted from water levels no influence on the FoS. in standpipes would result in non-conservative FoS When comparing the three different assumptions values if the failure plane would be above the tip of regarding the rate of pore pressure build-up with the standpipes. depth (Case A, B and C), there is no clear correlation The limitations of using standpipe piezometers for on which case results in the greatest difference in the determining the FoS of a TSF is evident, particularly FoS. It is only for the starter wall geometry and con- if the true pore pressures are unknown. sidering failure up to the first and second benches (Fig. 8 & Fig. 9) that there seems to be the greatest difference in the FoS for the two analysis method for Case B (lower rate of pore pressure increase at the toe) when the phreatic level is highest (Phreatic

439 17th ARC Conference 2019 7 RECOMMENDATIONS mechanism than the assumption made regarding the pore pressures. Firstly, it is recommended that the analysis conducted for this paper is considered as hypothetical only. The authors attempted to idealise the geometry and pore 8 ACKNOWLEDGMENTS pressure regimes from what has been observed from CPTu tests on TSFs in their experience. Since the FoS The authors would like to acknowledge the assistance in any stability analysis is sensitive to both geometry of Jones & Wagener in supplying the necessary soft- and effective stress (material strength and pore pres- ware to perform the analyses. sure dependent), each case of stability analysis on a TSF should be considered in isolation and therefore no generalisations from this article can be applied. 9 REFERENCES From the conclusions it is essential that understanding of the pore pressure regime is required be- Blight, G.E. 2010. Geotechnical engineering for mine waste fore analysis is conducted using water levels from storage facilities. Taylor & Francis. standpipe piezometers. Further, single standpipe pie- Cressie, N. 1990. The origins of kriging. Mathematical Geology. 22(3): 239-252. zometers are not sufficient for this and CPTu probing Morgenstern, N.R. & Price, V.E. 1965. The analysis of the sta- in the location of the standpipe would be required bility of general slip surfaces. Institute of Civil Engineers: to determine the pore pressure regime around the London. standpipe. Multiple standpipes installed to various Ridley, A. Brady, K.C. & Vaughan, P.R. 2003. Field measure- depths at a single location can also be used to infer ment of pore pressures. Transport Research Laboratory, TRL Report 555, Crowthorne. the rate of pore pressure build-up with depth at that Van der Berg, J.P. 1995. Monitoring of the phreatic surface in a location. tailings dam and subsequent stability implications. Masters The pore pressure regime can also be used to dic- dissertation. University of Pretoria. tate the depth to which standpipe piezometers must be Vermeulen, N.J. 2001. The composition and state of gold tail- installed. Van der Berg (1995) concluded that the FoS ings. PhD dissertation. University of Pretoria. from stability analysis conducted with piezometer Wagener, F.M. van der Berg, J.P. & Jacobsz, S.W. 1998. Moni- toring of the seepage regime in tailings dams. South African water levels can either be conservative or noncon- Institute of Civil Engineering Journal. 6(3): 23-26. servative, depending on the location of the critical failure plane in relation to the tip of the piezometers. Besides an understanding of the pore pressure regime in a TSF, prediction of the critical failure mechanism (or any number of failure mechanisms that need to be monitored in terms of the FoS against that mechanism being mobilised) should dictate the installation location and depth of standpipe piezometers. Should there be a change in the pore pressure regime, the failure mechanism of concern is likely to change, and the positions of standpipe piezometers may have to be evaluated and replaced. Trigger levels should also be determined for these piezometers which can indicate that the water levels have risen enough to warrant CPTu probing. In summary, standpipe piezometers must not be used in isolation for determining the stability of a TSF. The information must be supplemented by pore pressure data. These recommendations highlight the need to conduct regular CPTu testing to keep abreast of the changing pore pressure regime in a TSF if the stability of the TSF needs to be monitored. Finally, it is recommended that the stability analysis is determined at limit states (finite element analysis) for the two methods of prescribing pore pressures to the model so that the failure mechanism can be kin- ematically verified. Various foundation conditions should also be considered which might have a greater effect on the location and depth of the critical slip

440